Comparative Laboratory Studies of Conventional and Electrodynamic Fragmentation of an Industrial Mineral

The Optimized Comminution Sequence (OCS), an extensive laboratory experimental work scheme, has been developed at the Chair of Mineral Processing, Montanuniversität Leoben [18‐20], and is based on the idea of establishing a comminution process that minimizes the energy consumption ∆e for a given size reduction step i. The OCS is a tool developed to characterize comminution properties of minerals and rocks regarding their individual breakage characteristics independent from the influence of machinery. The specific energy consumption is a function of product dispersity, which is described by the particle size distribution and the specific surface area of the products [21, 22].

Numerous comminution stages in closed circuit, with pre-screening at small size-reduction ratio (<1:4) and high circulating load (>100%) are the characteristics for the OCS obeying the principle of energy-optimized comminution. The adjustment of the comminution tool of each stage is optimized with regard to the feed size. Fines in the feed are separated by pre-screening, directing the energy supplied by the comminution tool to the coarse particles. Each comminution cycle ends with intermediate classification at a defined screen aperture to remove the fine particles swiftly after their creation. This short retention time of the particles results in a smaller number of stress events per particle and cycle. Additionally, this leads to a material specific particle size distribution and minimum energy dissipation by compaction. The net energy consumption and the specific surface area of selected particle size classes of the comminution product are measured at each stage, as well as the particle size distribution of the feed and the comminution product [18, 21, 23].

When brittle mineral matter is fragmented in subsequent sub-circuits of small size reduction ratio, known as OCS, this results in the steepest possible cumulative fragmentation due to the lowest amount of fines, and minimum variation of particle sizes (and specific surfaces) possible at this particular maximum particle size. This is one aspect of the Natural Breakage Characteristics (NBC). After further comminution steps, the resulting fragmentation sieving curves plotted on a GGS-grid (Gates, Gaudin and Schuhmann plot) are shifted vertically upwards in a full logarithmic diagram, indicating self-similarity of the comminution progress. This implies that the local slope of the curve depends only on the “pure” fragmentation itself and is independent of equipment and processing [18, 20].

The cumulative display of the particle size distribution in the full logarithmic GGS-grid allows the distinction between homogeneous and inhomogeneous raw material behaviour by the degree of linearization of the distribution [18]. Particle size distributions of OCS products of inhomogeneous materials do not follow any power function, but the local uniformity factor (GGS exponent) remains largely independent from the maximum particle size. For homogeneous raw materials there is a linearization and ideally one single uniformity coefficient characterizes the total distribution over all maximum particle sizes. The particle size distributions from OCS are parallel to each other in the GGS-grid independent of raw material behaviour [22].

For an optimal sequence of fragmentation steps, a gain in mass specific surface (∆a, cm²/g), determined by permeametry, and specific energy consumption (∆e, J/g) are linearly dependent (down to a k_max of 40 µm). The slope represents a parameter of the rock, characterizing the natural fragmentation behaviour of the raw material, and its gradient is the Rittinger coefficient of comminution (R, cm²/J). According to [24], the Rittinger coefficient covers roughly a range between 10 cm²/J for low grindability samples (e. g. thermally compacted slags) and 150 cm²/J (e. g. unconsolidated limestone) [20, 25].

A complete characterization of the comminution behaviour of a given raw material is possible by the Rittinger coefficient, the uniformity coefficient, a raw material related shape factor and the nominal minimum particle size [23, 24].

Three Optimized Comminution Sequences are compared in this investigation. They consist of either one crushing (jaw crusher 1) or one electrodynamic stage (PWTS 100% or PWTS 250%), another crushing stage (jaw crusher 2) and finally two grinding stages (rod mill and ball mill) as demonstrated in Fig. 1. The different maximum feed size, split cut and circulating load of all stages are shown in Table 1. The circulating load given for comminution stage 1 is 100% for conventional comminution by jaw crusher 1. For the fragmentation and pre-weakening by high voltage pulses, in one sequence the circulating load was 100% (PWTS 100%), in the other 250% (PWTS 250%).

After each fragmentation step the comminution product was sieved manually, the oversize was topped up with fresh feed and recycled to the fragmentation unit, and the undersize was dried if necessary and stored for analysis. For the determination of particle size distribution, specific surface and energy consumption, either the mean was calculated from the three cycles in steady state for jaw crusher and mills, or representative samples were taken from steady state electrodynamic fragmentation. All the particle size distributions received from manual sieving were corrected by the results of air-jet sieving at 40 µm prior to the data evaluation and analysis.

The laboratory jaw crusher BB 200 by Retsch was used at the first stage (Jaw crusher 1 for the conventional comminution sequence only) and for the second comminution step during the OCS (Jaw crusher 2 for the conventional and both sets of high voltage treated process lines). The gap width was set to obtain the required product at a specific circulating load as given in Table 1. To measure and record the true power consumption in a standard electric circuit with symmetric load, a digital multimeter (VOLTCRAFT M‑4660M) connected to a PC by a RS 232-interface was used. The mean net energy consumption resulted from the total power draw reduced by the mean idle power draw measured before and after crushing, and the duration of the actual crushing action [23].

The electrodynamic fragmentation (and pre-weakening) by high voltage pulses was done with a pilot scale PWTS machine installed at SELFRAG in Kerzers, Switzerland. The PWTS unit presented in Fig. 2 comprises a pulse generator, a metal plate conveyor, a water vessel and a processing zone (Fig. 3). The immersed sample is processed between a disk-shaped top electrode which is connected to the pulse generator, and the flat bottom section of the metal plate conveyor acting as counter electrode. Adjustable parameter settings of the PWTS are voltage (50–200 kV), pulse energy (27.5–750.0 J), electrode gap (10–80 mm, which limits the maximum feed particle size), polarity (positive or negative), frequency (1–100 Hz), operation mode (batch with stationary or continuous with moving conveyor) and throughput (0–10 t/h). The installed power is 20 kW. The operational settings of the experimental work for PWTS 100% and PWTS 250% are shown in Table 2. The flexible generator setup with adjustable voltage and capacitance allows the testing with the same pulse energy, but at different voltages [12].

The used generator energy can be calculated for the PWTS from given parameter settings. It relates to the power taken from the power grid and refers to overall energy consumption, being a function of total capacitance of all capacitors used in the generator, the charging voltage supplied to the circuit, and the number of pulses applied. However, to date it is not possible to calculate or measure its actual spark energy like for the SELFRAG Lab, which is a laboratory unit of the same manufacturer. As the net fragmentation energy of the PWTS is not available, it is omitted at the first stage of all three comminution sequences. Instead, the focus is set to evaluate possible pre-weakening in subsequent fragmentation steps.

For the third comminution stage a rod mill, and for the fourth stage a ball mill were used. The dimensions and settings of both laboratory mills are given in Table 3 and are illustrated in detail in [22].

According to [19] there is, based on the principle of similarity, a relationship between the mass specific energy input for the dispersity step of rod mill (e_m, J/g), the inner diameter of the mill (d_i, m), the mass of grinding media (m_m, g), the fines generated (m_f, g), number of revolutions (n, 1), and gravity (g, m/s²) [22].

The dimensionless power conversion factor (or Steiner factor) c_p (1) for the mechanical power transferred into the grinding chamber was derived presuming a constant geometric and kinematic similarity and it is distributed in the range of 0.8–1.3 as determined from calibration measurements (laser aberration measurement) [19]. Bulk volume fraction of and friction conditions within the mill charge inside the mill chamber, the boundary between the charge and the shell of the tumbling mill and the ratio of centrifugal acceleration at the perimeter of the mill are expected to be rather constant [21].

Low differences between idle and load state, fluctuating voltage in the power grid and time-dependent changes in bearings and gears complicate the recording of electrical performance over time for a laboratory ball mill. An alternative approach is the measurement of the torque of the mill shaft using either a torquemeter with strain gauges, a pendulum torsion measurement, or a planetary gear. At the Chair of Mineral Processing, a torque sensor type (Hottinger Baldwin Messtechnik GmbH) with a measuring and a recording unit are used to record the energy consumption within the grinding chamber. The measuring shaft is bolted to the drive shaft between the electric motor and the mill by two rigid couplings. Four rotating strain gauges in a Wheatstone-bridge circuit are either compressed or stretched by the torque on the shaft. The measured signal is proportional to the torque. The reading in millivolt per volt of constant supply voltage is provided to a low voltage amplifier module QuantumX MX440B (HBM Inc.) which provides the time-dependent voltage signal to an electronic recording station. The software catman AP 3.4.2.52 (Hottinger Baldwin Messtechnik GmbH) allows an automatic conversion of the voltage signal into torque and calculation of its moving average [22].

The calculation of the mass specific net energy input (e_m, J/g) requires the amount of ground fines (m_f, g), the mean net torque (t_m, Nm) and the number of revolutions (n, 1) [22].

The evaluation of the specific surface area of all comminution products was based on permeametry. The arithmetic mean was calculated for three independent measurements of the surface with both a Blaine apparatus (by Tonindustrie Prüftechnik GmbH; constant volume) and the Permaran unit (by Outokumpu; constant pressure) for the fractions 0–40 µm and 40–100 µm [23, 25].

It can be assumed that the size classes 0–40 µm and 40–100 µm account for at least 90% of the total specific surface area of the samples and that the shape factor of the 40–100 µm fraction derived from surface equivalent particle size and the volume specific surface area of this particle size class is valid for all fractions >100 µm. Therefore it is possible to calculate their respective specific surfaces. Specific surface area of the entire comminution product is then calculated as a weighted average [21, 25, 26].

The feed was a common industrial mineral which had been crushed to 20–40 mm and washed prior to processing. Further details cannot be provided due to obligation of secrecy towards the industrial partner.